Last week I wrote a LinkedIn post on this topic and it quickly became my most-viewed post over the last year. Clearly, this challenge resonates with many marketers. That's why I've decided to expand that post into this tactical guide, detailing exactly how you can replicate this method for your customer research.
How can you create compelling GTM strategies if you don't truly understand your customers? As a PMM, customer insights are the foundation of effective messaging, but getting enough meaningful data points is sometimes very challenging.
Recently, I was tasked with creating targeted messaging for regions where I had limited data points. While I had a substantial internal perspective (product usage, web analytics, sales and customer service anecdotes), I lacked sufficient “voice of the customer”. I needed a quick, efficient method to gather customer insights from those specific regions.
The solution? This time I tried something new: a data-driven approach that combined review websites, SEO, and AI for deeper customer insights at scale.
In this post, I'll show you exactly how this method helped me analyze hundreds of authentic customer comments in hours rather than weeks.
The Challenge: Limited Customer Data Points
When I started this project, I relied on several “traditional” product marketing methods to gather customer insights:
Customer interviews
Internal team surveys
NPS scores and comments
Product usage data
Web traffic analysis
While these methods provided a foundation, I still faced significant gaps in understanding customers from regions where we had less direct interaction. I needed something more comprehensive and scalable.
The Results: Uncovering Hidden Customer Patterns
By combining an SEO crawling tool, review websites and ChatGPT, I discovered valuable patterns from customers in previously under-researched regions. These insights directly informed our messaging strategy, allowing us to speak more directly to the specific needs of these customers.
The best part? This process took hours instead of weeks and analyzed hundreds of customer comments rather than just a handful of interviews.
The Step-by-Step Process
Here's exactly how I combined these tools to uncover deeper customer insights at scale:
Step 1: Gather Raw Customer Commentary Using SEO Tools
I used Screaming Frog (an SEO spider tool) to crawl review platforms and gathered authentic, unfiltered feedback at scale.
Go to Configuration > Custom > Custom Extraction to select the elements you want crawled on the review pages. Use "Visual Extraction” and add a sample URL to find the exact section of the page you need to extract
Paste the URL of the review page you want to crawl (one or several)
Select the element on the page, in this case, the comment text, to indicate what you want to extract.
Pro tip: Most review platforms include metadata like user location/country and the date of the comment. Add these fields to your crawl —they'll allow you to segment insights by region and period later.
Crawl the pages containing reviews of your product or service
Export the crawled data to a spreadsheet
You could add as a tip to filter only the URLs needed, so they're not crawling the whole website.
Step 2: Prepare Your Data for Analysis
Before feeding the data to ChatGPT, I like to clean it up in a final Google spreadsheet, to make analysis easier:
Clean the spreadsheet by removing irrelevant columns
Ensure comment text is in a single, clearly labeled column
Use functions like =SPLIT() to extract text elements from a cell
Split contextual information like date, user location, and rating score if available in different cells (columns)
Remove any personally identifiable information to maintain privacy
Step 3: Leverage AI for Deeper Customer Insights
Back in the day, I used to use Pivot functions to try to make sense of large databases. Now, we have ChatGPT that can process a spreadsheet, add analytical insights in new columns, identify patterns, and summarize the findings, in a matter of seconds.
Here are a few examples of prompts that I used for this project:
Sentiment Analysis ChatGPT Prompt:
This spreadsheet contains customer comments about product X. Add a new column to the spreadsheet indicating whether column Y's text expresses a positive or negative sentiment. Positive sentiment includes phrases like "easy to use," "solved my problem," and "great customer service," while negative sentiment includes phrases like "difficult to navigate," "couldn't figure out," and "waste of time." Use these guidelines to categorize the text accordingly.
Note: The sentiment phrases are just examples. Use your own by cherry-picking examples from actual comments.
Feature Mention Analysis ChatGPT Prompt
Analyze column Z and add a new column after it that lists all product features mentioned in each comment. Use the following feature list as a reference: [list your key product features]. If a comment mentions a feature not on this list, include it as well. If a comment mentions a misspelled feature, include it as well.
Use Case Identification ChatGPT Prompt
Review column Y and add a new column after it that identifies the specific use case the customer is describing. Common use cases include: [list your known use cases]. If you identify a use case not listed here, note it as "New use case:" followed by a brief description. If you can’t identify any use case in the comment, note it as “No use case mentioned”.
Paint Point Identification ChatGPT Prompt
Review column Y and if it expresses a negative sentiment (includes phrases like "difficult to navigate," "couldn't figure out," and "waste of time”) add a new column after it that identifies the specific pain point the customer is describing. Common pain points include: [list your known pain points e.g. price too high]. If you identify a pain point not listed here, note it as "New pain point:" followed by a brief description.
Step 4: Analyze and Summarize Your Data
Once ChatGPT has enriched your dataset, you can start looking for patterns (this is where the old Pivot Table might still come in handy):
Pivot the data by region to identify location-specific trends e.g. top use cases by region, top product feature mentions by region, positive sentiment comments x use case by region etc.
Compare sentiment across different product features
Identify use cases and features that mostly drive the positive sentiments
Identify common pain points mentioned in negative reviews
Compare regional use cases and pain points to the ones you know in your main/most researched region
It’s really important to note that none of these insights should be taken at face value.
You should always leverage your cross-functional capabilities and validate your findings with internal sales and CS experts. If possible, see if the same findings are even partially validated by other data points like NPS comments, sales call insights (Gong has some really cool AI features for this), or product usage data.
Applying These Insights to Your GTM Strategy
The real value comes from putting these insights into action. Here's how I applied what I learned:
Refined messaging for specific regions based on the use cases and pain points most frequently mentioned in those areas
Created targeted content addressing the specific challenges faced by customers in different regions
Updated existing content to better address the needs of customers and prospects in the new regions
Conclusion: Scale Your Customer Research Without Scaling Your Time Investment
Through this approach, I was able to analyze hundreds of unfiltered customer comments in a fraction of the time it would have taken to conduct interviews. The insights proved invaluable for the new messaging strategy.
As marketers, we're always looking for ways to understand our customers better, but we don’t always have the time to conduct extensive interviews. This combination of ChatGPT, an SEO tool, and review websites offers precisely that balance—deep insights at scale.
Have you tried using SEO + GenAI tools like ChatGPT or Claude.ai to analyze customer feedback? I'd love to hear about your experiences. Connect with me on LinkedIn.
コメント